An Autonomous UAV Based Rail Tracking and Sleeper Inspection with Light-Weight Line Segmentation Approach

Abstract

With the development of technology in recent years, railway transportation is the most preferred transportation method in terms of comfort and safety. In railways, the sleeper component fixes the rail with ballast. Therefore, it is very important to determine the rail and sleeper problems that will affect the safety of the train during the operation of the railway system. In this study, an Unmanned Aerial Vehicle (UAV)-based rail tracking method is proposed for the control of railway track components and a method for inspecting the distance between sleepers by counting. The proposed method uses a lightweight deep learning-based line segmentation algorithm to detect rail and sleepers. By tracking the rail, sleeper counts are made from the images taken on the railway and the positions of the sleepers detected in the image are determined. Then, the distance between sequential sleepers is recorded as a time series, and anomalies in the time series and lost or shifted sleepers are detected.